VAGUES TECHNOLOGIQUES

PRODUIT INTÉRIEUR BRUT• Le PIB représente le résultat final de l'activité de production

des unités productrices résidentes d'un pays. Il peut se

calculer selon 3 approches :

1. La production : PIB = Valeurs ajoutées + Impôts sur la

production – Subventions .

2. Les revenus : PIB = Salaires + Excédant brut d’exploitation

des enterprises – Subventions + Impôts sur la production

3. La demande : PIB = Consommation Finale + Investissement +

(Exportations-Importations)

• Ces trois définitions doivent donner un résultat identique

INTRODUCTION

La croissance du PIB par tete est donc un phenomene recent

PIB Mondial est le meme en l’an zero et en l’an 1000

Taux de croissance moyen du PIBpartete mondial est de 1/19eme par an entre l’an1000 et l’an 1820

Ce taux de croissance monte a 0,5% entre 1820 et 1870, et depasse les 3-4% entre 1950 et 1973

EVOLUTION DE LA CROISSANCE

Crise de

Deuxième

Guerre

Mondiale

BERGEAUD-CETTE-LECAT

(2018)

INTRODUCTION

• Productivity Trends in Advanced Countries

between 1890 and 2012 , Bergeaud, Cette et Lecat,

Review of Income and Wealth (2016)

• Objectifs :

• Extraire de la courbe de croissance annuelle du PIB

par tête des tendances de long terme

• Comprendre ce qui sous-tend l’évolution du PIB par

tête en le décomposant

• Pays/regions consideres: Europe, USA, Japon,

Australie, etc.

DIFFUSION DES VAGUES

• En Europe, vague de croissance a lieu 20 ans plus tard, après la guerre

• Elle est stimulée par la reconstruction (habitations, machines, etc.) et la réorganisation des moyens de production : nouvelles technologies, nouvelles usines adaptées à l’électricité, nouvelles pratiques de management

• Elle est également soutenue financièrement par le plan Marshall

Croissance de la PGF (en %)

Zone Euro

Source : Bergeaud, Cette et Lecat (2016)

RÉVOLUTION NUMÉRIQUE

• Ralentissement de la productivité aux USA dans les années

1970 et 1980 (choc pétrolier et essoufflement des

innovations des années 1930)

• Puis nouvelle vague, dans les années 1990 jusqu’au début

des années 2000, avec la révolution des technologies de

l’information et de la communication (TIC)

Zone Euro

RÉVOLUTION NUMÉRIQUE• Toutefois, cette vague de croissance de la productivité est bien

plus faible que la précédente et presque inexistante en Europe

• Pourquoi pas de vague récente en Europe ?

Zone Euro

POURQUOI UN RETARD

DANS LA DIFFUSION DES

VAGUES

1) Importance des innovations secondaires

POURQUOI UN RETARD

VAGUES

2) Delai de diffusion dans les entreprises:

« la dynamo et l’ordinateur »

POURQUOI UN RETARD

VAGUES

3) L’amelioration dans l’utilisation des

technologies generiques

POURQUOI UN RETARD

VAGUES

4) L’adoption d’une nouvelle technologie

generique par les menages

POURQUOI UN RETARD

VAGUES

5) L’inadequation des infrastructures et

institutions locales

POURQUOI UN RETARD

VAGUES

6) Les limites de la mesure de productivite:

le cas de la revolution des TIC

CONCLUSIONS

• Au total, depuis 1820, le PIB par tête a connu une évolution

spectaculaire (take-off)

• Toutefois, cette évolution a été marquée par des phases de

hausse plus ou moins importantes

• Le moteur sous-jacent le plus important de cette croissance

du PIB par tête reste la productivité globale des facteurs, qui

est elle-même résultante de l’innovation, de l’éducation, etc.

• La diffusion des nouvelles technologies generiques tend a

etre retarde pour plusieurs raisons

What are the Labor and Product Market E�ects ofAutomation? New Evidence from France

Philippe Aghion, College de FranceCeline Antonin, Sciences-PoSimon Bunel, Insee and PSE

Xavier Jaravel, London School of Economics

October 5, 2020

Motivation

Tradeo�: productivity vs. displacement(e.g., Zeira 1998, Acemoglu-Restrepo 2019)

I Automation is labor-displacing at task level(e.g., Keynes 1930, Leontief 1952, Autor, Levy, and Murnane 2003)

I But could induce productivity gains, lower prices, higher demand, andneed for implementing new tasks(e.g., Autor 2015, Acemoglu-Restrepo 2018, Bessen 2018)

Several challenges when assessing this tradeo� empiricallyI Measurement of “automation”I Net e�ect likely depends on level of aggregation

This Paper

Contribute to literature in three waysI Study automation at plant, firm and industry levelsI Quasi-experimental variation from shift-share IV designsI Distributional e�ects: workers (employment) vs. consumers

(prices) vs. profits

Use linked employer-employee data set covering the Frenchmanufacturing sector (1994-2015)

Measure automation fromI (a) Balance sheet value of industrial machines at firm-levelI (b) Records of “motive power” at plant level, b/c common automation

technologies operate with electric motorsF Scope: not about AI/machine learning/computers

This Paper: Research Design

Descriptive evidence on population of firms and plantsI Event studies exploiting timing of adoption of industrial machines

across plants (in same firm) or across firms (in same industry)

Causal e�ect from firm-level shift-share research designsI Covers subset of firms which import industrial machines from abroadI Research design uses (a) changes in the productivity of foreign

suppliers of machines (“shifters”), along with (b) pre-determinedimporter-supplier relationships (“exposure shares”)

I Approximates an “ideal experiment” that would randomly assign priceof automation technologies across firms

Repeat shift-share design at the level of industries

Main Findings

Estimates indicate that increased automation leads to:I Increased plant-level, firm-level, and industry-level employment:

elasticity +0.3 after 5 yearsI Increased firm-level and industry-level sales: elasticity +0.4 after 5

yearsI Falling firm-level prices: elasticity -0.3 after 5 yearsI Also find increased profits, and cannot reject stable labor share at firm

and industry levels

Consistent with simple productivity - demand reallocation channel:I Automation ", marginal cost and prices #, demand ", employment "I Continues to operate at industry-level because of international trade

Roadmap

1 Data

2 Plant and Firm Level Analysis

3 Industry Level Analysis

4 Demand Reallocation Channel

Worker/Firm Data

Detailed information on workers and firms available from Frenchadministrative data (DADS and INSEE databases)

I Matched employer-employee data covering all plants in Frenchmanufacturing sector from 1994 to 2015

I Firms and plants: sales, industry, balance sheetI Workers: wages, occupation, education

Measuring Automation

Automation defined as “class of electro-mechanical devices that arerelatively self-operating after they have been set in motion on thebasis of predetermined instructions or procedures” (EncyclopaediaBritannica, 2015)

Firm-level: balance sheet value of industrial machinesPlant level: common automation technologies typically based onelectro-motive force, i.e. set in motion using electric motors

I Use plant-level records of peak electric capacity for motors (motivepower) used in production process

I Assembled by INSEE on large sample of plants; distinguishes betweenmotive power, thermic/thermodynamic uses, and other uses(electrolysis)

I Focus on motive power to exclude heating, cooling, servers

Motive Power Peak Capacity Measure

Advantages of motive power measure: available at plant level

Limitations:I Would not want to simply capture changes in electricity consumption

as a variable input ) to capture investments, use changes in “peakcapacity” for motive power

I Due to variation in e�ciency, di�cult to draw comparisons acrossindustries and over time ) analysis with industry/time fixed e�ects

Coverage

��

&KHPLFDOV 5XEEHU 3DSHU *ODVV�FHUDPLFV

)RRG�EHYHUDJHV

0RWLYH�)RUFH ,)5�5RERWV 3D\UROO 6DOHV

��

0RWRU�YHKLFOHV 5XEEHU 0DFKLQHU\ )RRG�EHYHUDJHV

0HWDO�SURGXFWV

0RWLYH�)RUFH ,)5�5RERWV 3D\UROO 6DOHV

Top 5 for Motive Force Top 5 for Robots

Chemicals

Rubber

Roadmap

1 Data

2 Plant and Firm Level AnalysisI Descriptive evidenceI Shift-share IV

3 Industry Level Analysis

Distributed Lead-Lag Model

Lit =10

d Lagk �Mi ,t�k +

Âk=�10

d Leadk �Mi ,t�k +µi +lst + eit

with employment Lit , change in electric motor consumption �Mi ,tand plant F.E. µi

Specification allows for delayed response of outcomes to increasedautomation

Causal interpretation requires E [�Mi ,t�k · eit |µi ,lst ] = 0 8 (t,k)I Pre-trends (leads) can be used as a falsification test - but cannot rule

out potential demand/supply shocks in contemporaneous periodI Mitigate potential correlated shocks with specifications using

industry-year or firm-year F.E., lst

Employment Dynamics

Start by documenting employment dynamics across across plants(with peak motive power measure)

Find that employment increases following increased automationI Elasticity of +0.2 on impactI Cumulative response increases further over time, with an elasticity of

+0.4 after 8 years

No pre-trends and magnitudes robust to changes in industry-year andfirm-year controls

I Implies potential confounding factors must have precisely the sametiming as automation and have stronger explanatory power thanfirm-year fixed e�ects (Oster 2019)

2-digit Industry by year F.E.

��

�(VWLP

DWHG�(ODVWLFLW\

�� <HDU�UHODWLYH�WR�FKDQJH�LQ

HOHFWULFLW\�FRQVXPSWLRQ�IRU�PRWRUV&RQWUROOLQJ�IRU��GLJLW�LQGXVWU\�E\�\HDU�)�(�

7RWDO�3ODQW�(PSOR\PHQW

4-digit Industry by year F.E

��

�(VWLP

DWHG�(ODVWLFLW\

Firm by year F.E.

��

�(VWLP

DWHG�(ODVWLFLW\

HOHFWULFLW\�FRQVXPSWLRQ�IRU�PRWRUV&RQWUROOLQJ�IRU�ILUP�E\�\HDU�)�(�

Employment Dynamics: Heterogeneity?

Are e�ects di�erent across skill groups?

I High/medium/low skills as in Charnoz-Orand 2017: non-production vs.skilled production vs. unskilled production workers

Find no heterogeneity across these skill groups

I Positive employment response for all, no change in relative wage

I Suggests no broad e�ect on inequality

I However heterogeneous e�ects could arise within skill groups,depending on set of tasks performed (consistent with Doms et al.1997, Hummels et al. 2014)

��

�(VWLP

DWHG�(ODVWLFLW\

3ODQW�(PSOR\PHQW��+LJK�6NLOO

��

�(VWLP

DWHG�(ODVWLFLW\

3ODQW�(PSOR\PHQW��0HGLXP�VNLOO

��

�(VWLP

DWHG�(ODVWLFLW\

�� <HDU�UHODWLYH�WR�FKDQJH

LQ�HOHFWULFLW\�FRQVXPSWLRQ�IRU�PRWRUV&RQWUROOLQJ�IRU��GLJLW�LQGXVWU\�E\�\HDU�)�(�

3ODQW�(PSOR\PHQW��/RZ�6NLOO

Robustness: Semi-Elasticity with Investment Event (> Threshold)

��

DWHG�6HP

L�(ODVWLFLW\

�� <HDU�UHODWLYH�WR�,QYHVWPHQW�(YHQW

7RWDO�)LUP�(PSOR\PHQW

��

DWHG�6HP

L�(ODVWLFLW\

�� <HDU�UHODWLYH�WR�,QYHVWPHQW�(YHQW

7RWDO�)LUP�(PSOR\PHQW

(a) > p75 (b) > p99More

Price Dynamics

Next, document the relationship between changes in automation andprice changes using export prices

I Export prices are measured as the unit value of exported products(available for all exporting firms from customs data)

��

DWHG�(ODVWLFLW\

�� <HDU�UHODWLYH�WR�,QYHVWPHQW�LQ�,QGXVWULDO�(TXLSPHQW

&RQWUROOLQJ�IRU��GLJLW�LQGXVWU\�E\�\HDU�)�(��DQG�+6��SURGXFW�)�(�

([SRUW�3ULFHV

[2/2] Shift-Share IV

Ideal experiment would randomly assign purchasing prices formachines across firms

Approximate with a shift-share research design, leveraging twocomponents:

1 Variation in the cost of imported machines over time acrossinternational trading partners (“shocks”)

2 Variation in pre-existing supplier relationships across French firms(“exposure shares”)

Intutively, French firms are di�erentially exposed to changes insector-specific foreign productivity

I Similar to Hummels et al. (2014), Amity et al. (2019), except that wefocus on machines

Shocks

“Shocks” are observed across trading partners by 4-digit industries:

I gn,t is change in imports flows of machines/robots from each tradingpartners (Germany, Italy, Japan, China, etc.) for each 4-digit industryacross 5-year periods

gnt =ImportMachinesn,t,t+5 � ImportMachinesn,t�5,tImportMachinesn,t,t+5 + ImportMachinesn,t�5,t

F where n indexes “trading partner by 4-digit industry” cells

I Similar result with leave-one-out SSIV (excluding focal firm)

Exposure Shares

“Exposure shares” of French firms:

I sint is share of trading partner n in firm i ’s total imports of machinesbetween t �5 and t �1

I Because of switching costs, French firm more likely to benefit from atrading partner’s productivity shock if it has a pre-existing importingrelationship with them

I Contemporaneous shares liable to reverse causality: use lagged shares

F Robustness: keep the shares fixed in the initial period, 1995-1999

Shift-Share IVConsider changes in employment �Lit and changes in motorconsumption �Mit over a five-year period across firmes indexed by i(omitting time subscript t)

We estimate by 2SLS:(�Lit = bZit + gXit + eit�Mit = aZit + egXit +eeit

with Zi the shift-share instrument constructed from shocks gn and(lagged) exposure shares sin � 0,

Zit =NÂn=1

sintgnt

Use panel with 5-year periods, 146 trading partners, and 239 4-digitindustries

Firm SSIV: First Stage

��

��\HDU�/RJ�&KDQJH�LQ�,QGXVWULDO

(TXLSP

HQW�IRU�/LQNHG�)LUPV

�� ,QVWUXPHQW��IRUHLJQ�VXSSOLHU�SURGXFWLYLW\�VKRFN�

Firm SSIV: Employment

�5 Employment

(1) (2) (3) (4) (5)

�5 Motor Cons. 0.307⇤⇤ 0.397⇤⇤ 0.444⇤⇤⇤ 0.411⇤⇤⇤ 0.430⇤⇤

(0.127) (0.153) (0.167) (0.149) (0.171)

First-stage F 14.39 11.58 8.526 12.08 7.697

4-digit Industry-year F.E. X X X X XLagged Firm Controls X X X XLagged Motor Cons. X XLagged Machines XContemporaneous Exports XN (trading parnter - industry - period) 7,250 7,250 7,250 7,250 7,250

Firm SSIV: Sales

�5 Sales

(1) (2) (3) (4) (5)

�5 Motor Cons. 0.395⇤⇤ 0.449⇤⇤ 0.512⇤⇤ 0.462⇤⇤ 0.461⇤⇤

(0.176) (0.208) (0.235) (0.206) (0.228)

First-stage F 14.39 11.58 8.526 12.08 7.697

Firm SSIV: Labor Share

�5 Labor Cost / Value Added

(1) (2) (3) (4) (5)

�5 Motor Cons. 0.00392 �0.0435 �0.0416 �0.0517 �0.0343(0.233) (0.230) (0.266) (0.226) (0.275)

First-stage F 14.39 11.58 8.526 12.08 7.697

Falsification Test: Lagged Firm Employment Growth

Lagged �5 Firm Employment

(1) (2) (3) (4) (5)

Instrument (gnt) 0.0111 0.0179 0.0167 0.0219 0.0162(0.0156) (0.0143) (0.0145) (0.0137) (0.0143)

Falsification Test: Lagged Firm Sales Growth

Lagged �5 Total Firm Sales

(1) (2) (3) (4) (5)

Instrument (gnt) 0.00354 0.0129 0.0117 0.0173 0.0111(0.0153) (0.0138) (0.0142) (0.0136) (0.0141)

Roadmap

1 Data

2 Plant and Firm Level Analysis

3 Industry Level AnalysisI Descriptive evidenceI Shift-share IV

[1/2] Descriptive Evidence

Positive plant-level and firm-level relationship between employmentand automation could in principle be overturned at the industry level

I Firms that automate less may be displaced by firm that automate more

Start with descriptive patterns: industry-level panel data with fixede�ects

I Patterns similar to firmsI Positive relationship with employment is stronger in industries more

exposed to international trade (ranked by export ratios)

Industry OLS Results: Employment

�5 Employment

Exposure to International Trade

All industries Above Median Below Median

(1) (2) (3) (4) (5) (6)

�5 Motor Cons. 0.406⇤⇤⇤ 0.275⇤⇤⇤ 0.401⇤⇤⇤ 0.382⇤⇤⇤ 0.111 0.016(0.065) (0.082) (0.124) (0.150) (0.082) (0.136)

2-digit industry by year F.E. X X X X X X�5 Other types of capital X X X

[2/2] Industry SSIV

Repeat previous SSIV, except that i is now a 4-digit industry (ratherthan a firm) and n is a “trading partner by 3-digit industry” cell

gnt =ImportMachinesn,t,t+5 � ImportMachinesn,t�5,tImportMachinesn,t,t+5 + ImportMachinesn,t�5,t

Measure these shocks across 179 trading partners in the 95 3-digitindustries during three 5-year periods, from 2000 to 2015

Examine the response of employment and sales in industries thatsource their machines from increasingly productive foreign suppliers

I For leave-one-out analysis, need to work with “trading partner by2-digit industry” and 3-year periods to preserve power

Industry SSIV: Employment

�5 Employment

(1) (2) (3) (4)

�5 Motor Cons. 0.558⇤⇤⇤ 0.584⇤⇤⇤ 0.589⇤⇤⇤ 0.620⇤⇤⇤

(0.193) (0.174) (0.184) (0.196)

First-stage F 23.52 14.43 13.66 11.67

3-digit Industry-year F.E. X X X XLagged Industry Controls X X XLagged Machines X XLagged Other Capital XN (trading partner - industry - period) 7,482 7,482 7,482 7,482

Industry SSIV: Sales

�5 Sales

(1) (2) (3) (4)

�5 Motor Cons. 0.486⇤⇤⇤ 0.585⇤⇤⇤ 0.594⇤⇤⇤ 0.627⇤⇤⇤

(0.169) (0.186) (0.199) (0.207)

First-stage F 23.52 14.43 13.66 11.67

Industry SSIV: Labor Share

�5 Labor Cost / Value Added

(1) (2) (3) (4)

�5 Motor Cons. �0.0311 �0.0286 �0.0285 �0.0282(0.0768) (0.0603) (0.0610) (0.0671)

First-stage F 23.52 14.43 13.66 11.67

Industry SSIV: Accounting Profits

�5 Profits

(1) (2) (3) (4)

�5 Motor Cons. 0.787⇤⇤⇤ 0.752⇤⇤⇤ 0.761⇤⇤⇤ 0.834⇤⇤⇤

(0.255) (0.222) (0.234) (0.234)

First-stage F 23.52 14.43 13.66 11.67

Falsification Test: Lagged Industry Employment Growth

Lagged �5 Industry Employment

(1) (2) (3) (4)

Instrument (gnt) 0.0313 0.0209 0.0207 0.0203(0.0282) (0.0173) (0.0166) (0.0165)

Falsification Test: Lagged Industry Sales Growth

Lagged �5 Total Industry Sales

(1) (2) (3) (4)

Instrument (gnt) �0.00807 �0.00559 �0.00496 �0.00309(0.0253) (0.0179) (0.0177) (0.0178)

Conclusion

Contribute to literature in three waysI Study automation at plant, firm and industry levelsI Causal identification using a shift-share designI Document impacts on workers (employment), consumers (prices)

and profits

Estimates indicate that automation can increase labor demand andcan generate productivity gains that are broadly shared acrossworkers, consumers and firm owners

In a globalized world, attempts to curb domestic automation inorder to protect domestic employment may be self-defeatingdue to foreign competition.

VAGUES TECHNOLOGIQUES

Documents

Transcript of VAGUES TECHNOLOGIQUES

Vagues impressionnantes

Considérations technologiques

Vagues des marees 2014 malouine

Terreur moderne : les quatre vagues

Interprété par Jacques Brel - 1962 Avec la mer du Nord pour dernier terrain vague, Et des vagues de dunes pour arrêter les vagues;

2 vagues COVID à Bourg-en-Bresse

Fnac Nouveautés technologiques

Louis-Ferdinand CÉLINE - DES VAGUES (1917)

DETERMINANTS TECHNOLOGIQUES ET NON TECHNOLOGIQUES´ …

cleantuesday Prototype HACE energie des vagues

avec innovations technologiques

Les Vagues - yourcenariana.org

Evolutions technologiques

Tubes Vagues

Prévoir les vagues et les interactions courants-vagues … · Zone de génération des vagues arrivant en Bretagne et ... S.I. = erreur en moyenne quadratique (RMSE) normalisée

Métiers & Compétences technologiques

ENSEIGNEMENTS TECHNOLOGIQUES TRANSVERSAUX

LES VOIES TECHNOLOGIQUES Les séries technologiques Les séries technologiques sont polyvalentes, pour permettre des poursuites d’études supérieures (BTS,

Guide saison Nouvelles vagues

formations industries technologiques